On Friday, 10 July 2015 at 20:38:16 UTC, Rishub Nagpal wrote:
On Thursday, 9 July 2015 at 06:32:28 UTC, Dan Olson wrote:
I've made a set of binaries and universal libs for the LDC iOS
cross-compiler. It is based on LDC 0.15.1 (2.066) and LLVM
Only 32-bit devices currently; arm64 work starts next month
when I acquire an iPhone 6.
The download should have everything needed to run on an OS X
build host in the same fashion as LDC downloads. But I may
have missed something. Feedback appreciated.
Good Work! I'd like to help get D to work on android, but I do
not know much about llvm and arm compilers to be of much help.
Last I heard there was an issue with exception handling and
TLS, is that still so?
That's funny, because I was just thinking about putting my
Android patches for ldc online and trying to get more people to
chip in on working through the remaining tests to be fixed for
I got TLS working a month and a half ago
(http://forum.dlang.org/post/imkgasjuvbbasyghd...@forum.dlang.org) and exception-handling seems to be working since this fix I ferreted out last week (http://forum.dlang.org/post/qsfaussopqwwjuljd...@forum.dlang.org). Now it's just codegen issues, with about half of phobos modules' tests failing somewhere, though many of those modules only have a handful of tests that fail. For example, only three unit test blocks fail in std.stdio and one in std.path. Common causes appear to be problems with ranges and functions from std.random.
I'm going through each module and commenting out failing tests
and checking backtraces, a time-consuming process that's got me
thinking about hacking the test runner, so that failing tests in
one unit test block won't stop other test blocks from the same
module from running, as is the case now.
If you or anybody else is interested in chipping in, reply in the
Android thread (first link above) and I'll put some patches and
build info online. Unfortunately, to really fix any of these
issues, you'll probably have to know something about ARM
assembly, LLVM IR, and be comfortable stepping through the binary
with gdb but without debug info, although simply triaging the
tests to figure out what works and what doesn't could probably be
done by almost anyone.